New Google Gemini Update Is INSANE!
The video showcases seven practical use cases of Google's new Gemini Flash model, demonstrating its ability to analyze images, process long videos, review contracts, generate code, transcribe audio, debug errors, and operate across Google apps. The presenter argues that Gemini Flash outperforms the larger Pro model and highlights a hidden settings tweak that improves answer quality. The content is framed around real business workflows used for the presenter's AI Profit Boardroom community.
Summary
The video opens with the presenter noting that Google unexpectedly pulled Gemini Pro from center stage and promoted Gemini Flash — the smaller, faster model — which is reportedly outperforming its larger counterpart. The presenter, a digital avatar representing Julian Goldie (CEO of Goldie Agency), claims to have tested the model extensively on real business tasks throughout the week.
The first trick involves dropping a photo of a messy, coffee-stained whiteboard into Gemini and asking it to generate a clean growth plan. The model reportedly read partially obscured words correctly and returned a structured plan with identified gaps, a task the presenter says used to take an entire morning.
The second trick demonstrates Gemini's long-video comprehension. A full training session was uploaded, and the model returned accurate timestamps for the top five lessons and even reconstructed a chart from the video — all without opening any external tools. The presenter suggests this capability can turn a single training session into a full week's worth of community content.
Trick three covers contract review. A 50-page partnership agreement was analyzed using the 'thinking level' setting. On low, the model gave a basic summary, but on high, it caught two penalty clauses and a hidden auto-renewal that the low setting missed entirely. The presenter frames this as a rule: use high thinking level when mistakes carry real consequences.
Trick four involves building a working landing page from a hand-drawn sketch. The presenter photographed a rough box-and-line drawing, prompted Gemini to build a landing page for the AI Profit Boardroom, and received hundreds of lines of functional code. When pasted into a live preview, the buttons and layout worked correctly — described as going from idea to clickable product in minutes.
Trick five is a daily productivity habit: recording a 2-minute voice note of unstructured morning thoughts and dropping the audio into Gemini, which transcribes, groups, and prioritizes tasks automatically before the workday begins.
Trick six addresses workflow debugging. A screenshot of a confusing automation error was submitted, and Gemini identified the root cause in plain English and provided a step-by-step fix, resolving the issue in minutes rather than hours of forum searching.
The seventh and final trick involves enabling Google extensions (Drive, Docs, Gmail) and using a single prompt to find a file in Drive, create a summary document, and draft an email with a link to that doc — all executed autonomously across three apps without the user touching any of them.
The video closes with an important settings tip: when users migrated to the new Gemini model, the thinking level silently defaulted from high to medium, degrading answer quality without any announcement from Google. The fix is to manually switch it back to high in settings, which the presenter says takes 15 seconds and immediately improves outputs.
Key Insights
- The presenter claims Google's smaller Gemini Flash model is outperforming the larger Gemini Pro model, and that Google quietly prioritized Flash after pulling Pro from its launch spotlight.
- The presenter found that switching Gemini's 'thinking level' from low to high caused it to catch two penalty clauses and a hidden auto-renewal in a 50-page contract that the low setting missed entirely — framing the setting as a risk management tool.
- The presenter reports that when users migrated to the new Gemini model, the thinking level silently defaulted from high to medium without any announcement from Google, meaning many users are receiving degraded answers without knowing it.
- The presenter demonstrated that Gemini can autonomously operate across Google Drive, Docs, and Gmail from a single prompt — finding a file, creating a summary document, and drafting an email with a link — without the user interacting with any of those apps directly.
- The presenter argues that Gemini's long-video analysis capability — which watched a full training session, returned accurate timestamps, and reconstructed a chart — can turn a single recorded training into a full week of community content without any additional tools.
Topics
Full transcript available for MurmurCast members
Sign Up to Access